[GitHub] [hudi] alexeykudinkin commented on a diff in pull request #6815: [HUDI-4937][Stacked on 7702] Fix `HoodieTable` injecting non-reusable `HoodieBackedTableMetadata` aggressively flushing MT readers

["alexeykudinkin (via GitHub)" <gi...@apache.org>]

alexeykudinkin commented on code in PR #6815:
URL: https://github.com/apache/hudi/pull/6815#discussion_r1083175593


##########
hudi-common/src/main/java/org/apache/hudi/metadata/BaseTableMetadata.java:
##########
@@ -73,8 +73,10 @@ public abstract class BaseTableMetadata implements HoodieTableMetadata {
 
   private static final Logger LOG = LogManager.getLogger(BaseTableMetadata.class);
 
-  public static final long MAX_MEMORY_SIZE_IN_BYTES = 1024 * 1024 * 1024;
-  public static final int BUFFER_SIZE = 10 * 1024 * 1024;
+  protected static final long MAX_MEMORY_SIZE_IN_BYTES = 1024 * 1024 * 1024;
+  // NOTE: Buffer-size is deliberately set pretty low, since MT internally is relying
+  //       on HFile (serving as persisted binary key-value mapping) to do caching
+  protected static final int BUFFER_SIZE = 10 * 1024; // 10Kb

Review Comment:
   What we prefetch here isn't actually used past parsing the block header (magic, version, etc) so there's no point in fetching large slabs. The reason why this is not going to be used is b/c we use HFile reader which does its own caching of the blocks



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@hudi.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org